Considering our findings of a negative effect of a high PGR level on PC outcome in the first paper, we sought to further elucidate the significance of PGR in PC. This time we
systematically assessed the two receptor isoforms, PGRA and PGRB, their stromal and epithelial distribution and association with clinical outcome. In addition, the
clinicopathological data in our cohort had recently been updated, rendering longer follow-up time and more events. The receptors were otherwise investigated in the same manner as our previous markers.
4.4.1 Receptor expression
PGRA expression was detected exclusively in stromal cells in both normal and malignant tissue. Expression of PGRB was both stromal and epithelial, and PGRB was located in all tissue compartments. The expression of PGRA in stromal cells was significantly higher compared to PGRB in both NS (p < 0.001) and TS (p < 0.001). Regarding PGRB, the epithelial expression was overall higher than the expression in the surrounding stroma (p <
- 86 -
0.001). Finally, a strong and significant correlation between PGRB expression in TE and TS was detected (r = 0.82, p < 0.001).
4.4.2 Univariate analyzes
Patients with a high density of PGRB had a significant decrease in both BFFS and CFFS. This applied to both TE expression (BFFS: p < 0.001, CFFS: p = 0.006) and TS expression (BFFS:
p = 0.034, CFFS: p = 0.034). No additional prognostic value was evident when merging PGRB expression in TE and TS. No association with clinical endpoints was discovered for PGRA expression.
4.4.3 Multivariate analyzes
A high density of PGRB expression in TE remained an independent prognostic marker for both BF (HR: 2.0, 95% CI: 1.45 – 2.76, p < 0.001) and CF (HR: 2.5, 95% CI: 1.29 – 4.85, p = 0.006). Non-apical PSM (p = 0.016), Gleason grade group 3 and 4 (p = 0.032 and p = 0.008), PNI (p = 0.002), preoperative PSA (p = 0.021) and pT-stage 3b (p = 0.001) were additional independent prognosticators for BF. Regarding CF, age ≥ 60 (p = 0.026), LVI (p = 0.028) and Gleason grade group 1 through 5 (p = 0.013) were additional independent prognosticators.
Patients with a high PGRB level in TE had twice the risk of experiencing BF and 2.5 times the risk of CF compared to patients with low levels. PGRB in TS did not reach statistical significance in multivariate analyzes.
- 87 - 4.5 Overview of the main results
Table 8 - Overview of main results from uni- and multivariate analysis in paper I – III. (Univariate analyzes:
log rank test, multivariate analyzes: Cox regression analyzes, backwards stepwise model) Abbreviations: TE = Tumor epithelial cells; TS = Tumor associated stromal cells; PGR = Progesterone receptor; ER = Estrogen receptor; BF = Biochemical failure; CF = Clinical failure; PCD = PC death; HR = Hazard ratio; ns = Not significant; gray shaded square
= not entered in analysis
- 88 - 5 DISCUSSION
5.1 Patient cohort
A strength of this thesis is the unselected study population from Central and Northern Norway. Although this is a relatively small populace, several aspects make this a
heterogeneously composed cohort. The median life expectancy for men in Norway was 80.6 years in 2015. In the study population (Paper I, II and III) it ranges from one of the lowest estimates of 79.0 years in Nordland to one of the highest with 80.6 years in More and
Romsdal. Life quality surveys have also revealed a broad specter the regarding percentage of population classifying as obese (BMI ³ 27), ranging from 22 % to 35 % with some of the counties well above the national average. Further, the part of the population who never exercise varies from 12 % – 17 % depending on the county, and from 10 % to 16 % of the men are daily or “some-times” smokers (www.SSB.no, numbers from 2015 – 2016). These factors are all variables associated with PC, however the influence on cancer progression cannot fully be determined28.
A reasonable estimate, according to the head of the department of Urology, UNN Tromso, T.
Knudsen, is that ³ 95 % of men diagnosed with PC in these regions during the inclusion period were operated at the hospitals participating in this study. This thesis is based on a large PC cohort of 535 patients. We did, however, encounter a potential selection bias when
collecting PC tissue for the cohort from St. Olav´s hospital. A great number of PC specimens from this hospital were appropriated by another research group (n = 100) and not available at the time of tissue collection. Thus, a part of the St. Olav´s cohort could not be included in our material, reducing the total cohort size and representability.
A retrospective study is less expensive and more time efficient compared to e.g. prospective studies. The retrospective design has its natural limits in the access to additional information regarding e.g. life style and comorbidities. It also removes the option of collecting additional material that could be of interest, e.g. blood samples. Additionally, it always renders the possibility of information bias as we have limited ways of verify the information
retrospectively collected from patient journals. The retrospective design also excludes the possibility to standardize follow-up procedures. For example; if one center uniformly measured PSA levels every 3 months they could have more BF events and shorter BFFS.
Another bias occurs in studies that includes a mixture of PC material collected before and
- 89 -
after the introduction of the PSA-test. Introduction of the PSA-test has resulted in increased detection of PC with indolent disease10. Our material is collected between 1995 – 2005, and thus placed in the PSA-area. This renders the assumption of a more homogenous and comparable material.
When tumor material has been collected over a long period of time, a challenge resides in alterations to the diagnostic guidelines and procedures. This could affect the tumor types included in the material and results in variations in cancer staging. This potential bias is evaded in our material by the re-classification of tumors according to updated diagnostic systems. This material was further collected prior to the introduction of image guided biopsies296. It is thus reasonable to assume that the diagnostic procedures for detecting PCs were uniform with the standard DRE, PSA testing, and TRUS with needle biopsies69. This thesis benefits from a material with a long follow-up time. Due to the nature of PC, the number of PC specific deaths remians low despite decades of follow up. By evaluating other endpoints associated with disease progression (BF, CF)89,109, in addition to PCD, more robust statistical results could be produced due to a greater number of events. It must, however, be noted that, although BF is a early sign of disease activity, the time to progression to CF varies to a great extent. Further, not every patient with BF will experience CF and PCD within their lifetime89,109,111. The international definition of BF is currently two consecutive postoperative rises in PSA level > 0.2 ng/mL106. This is, however, a topic of controversy. There are several arguing that that a higher PSA level cut-off of ³ 0.4 ng/mL is stronger associated with continued systemic progression, and consequently makes a more clinically relevant cut-off104,105. Based on this assumption, we chose to establish cut-off for BF at 0.4 ng/mL to ascertain that the patients identified where those at high risk of clinical progression. Whether to evaluate disease specific survival or overall survival is another consideration. PC affects older men and, for majority of cases, the cancer progression is slow4. Comorbidity and death from other causes than PC is prevalent; disease specific survival is therefore the chosen parameter herein. Disease specific survival is further dependent on accurate entries regarding the cause of death, this can be biased by subjective interpretation.
A confounder when performing survival analyzes on material with a long follow-up time is changes in post-operative treatment regimes. Benefitting our material, the standardized
- 90 -
treatment strategies and the equality in health care distribution in Norway enables a relatively homogenous study population. However, with the introduction of new treatment strategies, such as new generation hormonal therapies243,194 and improved bone targeted terapy244, great advances have been made in treatment of castrate resistant- and metastatic PC. This has led to an increase in survival rates over the past years2 and can consequently affect the results regarding impact of different molecular markers on disease specific survival. On the other hand, these new treatment strategies have little impact on other clinical endpoints such as BF and CF.
In this material, the exclusion criteria I and II were included to minimize bias introduced by mechanisms that may alter the TME in a manner not related to PC biology. Radiation therapy might induce necrosis or alter the protein structure in the tumor tissue. Previous malignancies can change the host’s biological response to the current malignancy, its treatment could have affected the PC tissue and metastatic disease could be misinterpreted to represent the wrong primary cancer. Criteria III represents the greatest number of excluded patient and is
important to minimize bias related to tissue processing and analyzes.
5.2 Methodological considerations 5.2.1 Tissue fixation and processing
A great benefit of FFPE tissue is that it eliminates the need for fresh or fresh – frozen tissue, it conserves tissue morphology, and it can be stored for many years and still exhibit stabile immunostaining for most antigens297,298. Our material was collected over a ten – year period and thus using FFPE was the most applicable preservation method. Cutting sections from TMA can be technically more challenging than cutting whole tissue sections, making TMA more prone to certain artifacts and tissue loss. FFPE preserved over many years can lose its elasticity, which would lead to challenges in “punching out” i.e. obtaining cylindrical cores for constructing TMAs299. This resulted in a number of “missing” cores in our cohort, which was somehow higher than expected.
In theory, every step in the pre-analytic phase (Figure 22) has several variables that challenge standardization, which has been thoroughly reviewed by Engel & Moore278. A major
challenge with formalin fixation has been alterations of the tissue proteins three-dimensional structure (e.g. cross-linking of proteins and DNA), thereby masking or damaging epitopes277.
- 91 -
Today, this has been improved by antigen retrieval methods. This technique increases the accessibility to tissue antigens, and was considered a breakthrough for IHC based research when developed in 1991286. As reviewed by Engel & Moore, formalin fixation is both a time-dependent and time-consuming method. If the tissue has been placed in formalin for a long time ( > 24 – 48 hours), it can result in “over-fixation” which can weaken or destroy the tissue antigenicity and result in a false negative staining278. The development of antigen retrieval methods has improved the epitope detection following prolonged tissue fixation286, but standardizations in tissue fixation times are lacking. Formalin fixation has an average tissue penetration time of 1 mm per hour277. Thus, the optimal fixation time can vary depending on multiple factors, including the size and the consistency of the specimen. Consequently, longer fixation is needed for larger tissue samples to prevent autolysis and deterioration of tissue antigenicity300,301. There are also several aspects of tissue processing techniques that can compromise antigenicity. For instance, non-specific staining can occur from inadequate tissue dehydration protocols and by the use of dehydration reagents of insufficient quality prior to paraffin embedding. Additionally, differences in temperatures during paraffin embedding can also affect IHC quality278. Given that our cohort was collected retrospectively from several pathology institutions over a time period of 10 years, variations in tissue fixation was expected and must be taken into account. Additional quality assessments were made by stratifying the significant results based on the different donor institutions and 5-year time intervals. In our studies, the same trends in results were observed throughout time and
pathological centers (Paper I, II and III), although not always with significant results in each subgroup. This could probably be due to the reduced number of patients and the reduced number of events in each subgroup.
5.2.2 Tissue microarray procedure
A major benefit with TMAs is the tissue utilization, in addition to the time and
cost-effectiveness. A single TMA experiment can provide data on the molecular characteristics of as many as 1000 specimens at once279. This also minimizes experimental variability since the same experimental conditions are applied to all tissue specimens on one master-slide
simultaneously, increasing the reproducibility of the staining reactions. In contrast, the conventional whole section analyzes will contain only one tissue sample. As a result, up to a thousand separate analyzes, with a new experimental procedure for each slide, is required to obtain the same results as for a single TMA slide. Given the frequent limitation in tissue
- 92 -
availability for research purposes, TMAs can multiply the number of studies one tissue sample can supply. TMAs also provides the possibility of assembling a TMA panel with acquired healthy tissue samples, which is a great benefit. By doing this, controls tissues will be readily available when needed for different laboratory procedures, e.g. antibody validation.
Having said this, we are aware of that malignant tumors can be heterogeneous in general, and that prostate cancer is in particular known to show a varying grade of heterogeneity48–52. Typically, the tumor cores in PC are sampled from the most dominant tumor (index tumor).
However, observations have been made of metastatic tissue not sharing genetic changes with the index tumor and of non-dominant tumor foci giving rise to metastasis54,55. This can present an even bigger challenge when using TMA as a method, and there has been concerns regarding the representatively of this method compared to the conventional whole tissue sections302,303. However, several studies have demonstrated a good reproducibility of biomarker expression in TMAs compared with whole section tissue analyzes304,305 and clinical endpoints306, including PC307. An overview of advantages and weaknesses with TMA is presented in Figure 23.
Several of the initial TMA studies were conducted on FFPE PCs. As a result, throughout the years, knowledge on how to apply TMA based PC research has been optimized. It continues to day to serve as an acknowledged method for gene and protein detection in PC308. In several tissues, two or more cores have been found to supply satisfactory information regarding antigen prevalence adequately to a whole tissue sections304,309. However, three to four cores are needed to optimally investigate biomarkers expression in PCs due to the aforementioned tumor heterogeneity310. Importantly, using larger cohorts, like ours, one would additionally benefit from dilution of potential sampling errors302. In our cohort, an average of four cores per patient were collected, with an average of two of four being cores from index tumor.
Additionally, the cores were selected by an experienced pathologist, which is essential to ascertain tissue representativeness. Our TMA sections were cut 4µm thick, as is
recommended, given that thicker sections can result in increased background staining276. Inevitably, tissue morphology will change through the length of the tissue core as more TMA slides are cut. To prevent incorrect classification of tumor areas, repeated H&E re-staining and re-evaluations of the sections (every 50th section or so) is recommended. The TMA slides applied in this thesis were cut early in the process and were thus affected to a minor degree by
- 93 -
this bias. Finally, when FFPE material is cut and affixed to glass slides the antigen epitopes are exposed and vulnerable, thus if the IHC process is delayed the quality could deteriorate299. Being conscious of this, either IHC was preformed within a day, or the slides were prepared for storage by sealing them with paraffin and then kept in refrigerator storage at
approximately + 4°C for no more than one year.
Figure 23 – Advantages and weaknesses of tissue microarray. Figure by Thea Grindstad
5.2.3 Antibodies
Regarding antibodies, the objective is to achieve a strong, specific antigen-antibody signal with minimal background staining. Although the use of IHCs is extensive, a generalization of the selection and validation process is lacking311. Today, antibodies are commercially
available from numerous manufacturers, automatically introducing challenges as price and quality. Additionally, different antibodies are suited for different tissues and procedures, representing a selection bias. Insufficient antibody selection and validation has always been, and continues to be, a common concern312,313. Binding of an antibody to an epitope is a reversible process and depends on precise antibody – antigen interaction and their binding affinity291. Changes in antigen conformation can affect the strength of this interaction, e.g. the challenge with formalin tissue fixation278. Antibodies cross-reacting with similar epitopes on different antigens and background staining, due to hydrophobic and ionic interactions and
Time/ cost saving and high throughput - Reduces number of experiments per cohort - Less amount of reagent is required
- Expands tissue use of limited tissue resources
- Analyzes multiple cases per slide (£1000) Reduces experimental variability
- Uniform reaction conditions -Accesible control material -A population-level research tool
- Tissue morphology can change through the length of a tissue core
- TMA is not valid for individual diagnosis - TMA construction requires techincal skills and special equipment
- Securing representative tissue sections and sufficient tissue heterogenity
- Representability of the total tissue core remains debated
Advant ages We ak ne ss es
- 94 -
endogenous enzyme activity represents additional challenges283,291. We met these challenges by carefully selecting antibodies based on experience, availability and expertise, and by always applying recommended positive and negative controls and non-IHC validation methods when deemed necessary. The applied pan-PGR and PGRA antibodies have been extensively validated and applied in routine breast cancer diagnostics199,314. Pan-PGR, ERa, ERb, and aromatase were subjected to further in-house validation. Additionally, all applied antibodies were validated by the manufacturer. Despite these precautions, the validity of our ERb antibody (clone PPG5/10) has been questioned in recent publications312,313. This antibody is directed against the ERb1 isoform, the wild type ERb, and is one of the most widely applied ERb antibodies today.
In this thesis, both monoclonal and polyclonal antibodies were applied: Monoclonal: ERb (Paper II), PGRA and PGRB (Paper III). Polyclonal: pan-PGR (Paper I), ERa and
aromatase (Paper II). Each have their own benefits and disadvantages (Figure 24). Polyclonal antibodies are technically easier, less expensive and less time consuming to produce
compared to monoclonal antibodies291. They also have higher detection sensitivity and are stable in a broader range of conditions compared to monoclonal antibodies. They are,
however, also at greater risk of creating false positive results due to cross-reactivity with other epitopes291. Monoclonal antibodies are highly specific by binding to one single epitope on an antigen, thus in less risk of cross-reactivity292. They are however less stable and threatened by conformational changes to a larger extent than polyclonal antibodies. This is given that the detection depends on only one epitope which increases the risk of false negative results.
Monoclonal antibodies are also a constant renewable source when generated, whilst production of polyclonal antibodies is limited to the applied animal and its lifespan291.
- 95 -
Figure 24 – The main differences of monoclonal and polyclonal antibodies summarized. Figure by Thea Grindstad
5.2.4 Immunohistochemistry procedure
IHC is a well-established method for in situ evaluation of the prevalence of various antigens, as well as their localization and distribution in different tissue compartments. It can be performed on both fresh tissue samples and on fixed tissue, including both small (e.g. TMA, biopsies) and whole mount sections. Herein (Paper I, II and III), an indirect IHC technique was applied which is a more sensitive method of antigen detection compared to the direct IHC. This is due to the signal amplification provided by the secondary labeled antibody, which makes it well suited for detection of infrequent and heterogeneously expressed
antibodies. It does, however, require several additional stages of incubation283. Direct IHC is more applicable for the detection of highly expressed antigens, and represents a less
demanding methodology283. Indirect IHC as a procedure harbor several steps (Figure 21)
demanding methodology283. Indirect IHC as a procedure harbor several steps (Figure 21)